论文详情
地震相是沉积相在地震剖面上的反映,能为地下资源尤其是油气资源的勘探开发提供有效依据。近年来随着人工智能的快速发展和油气人工智能的有力推进,国内外学者提出了多种地震相智能识别的方法。对地震相智能识别方法进行了归纳总结,将其归纳为无监督学习、监督学习和半监督学习3类,并详细介绍了这3类方法的原理、应用现状及其优缺点。无监督学习利用没有标签的地震数据进行学习聚类,从而实现地震相的自动识别,具有简单易操作的特点。监督学习主要利用标签数据反馈学习,通过学习不断接近标签,从而使得该方法在地震相识别中具有更高的精度。半监督学习在地震数据标签不足的情况下,利用合成伪标签等方式进行学习,但伪标签中存在的误差会降低该方法的精度。最后以神经网络地震相识别为例,对地震相智能识别技术进行了展望。
Seismic facies are the reflection of sedimentary facies on the seismic profile,which can provide a favorable basis for the exploration and development of underground resources,especially for oil and gas.In recent years,with the rapid development of artificial intelligence and its vigorous advancement in oil and gas,researchers have proposed several methods for intelligent identification of seismic facies.This article summarizes the intelligent identification methods for seismic facies and classifies them into three categories:unsupervised,supervised,and semi-supervised learnings.Furthermore,it introduces the principles,application status,advantages,and disadvantages of these three categories in detail.Unsupervised learning uses unlabeled seismic data for learning clustering to realize automatic identification of seismic facies,which is simple and easy to operate.Supervised learning mainly adopts label data to feedback learning and continuously approaches labels through learning,so that it has high accuracy in seismic phase recognition.Semi-supervised learning applies synthetic pseudo-labels and other methods for learning when there are insufficient seismic data tags,but the errors in the pseudo-labels reduce its accuracy.Finally,an example of neural network seismic facies recognition is shown,and the intelligent seismic facies recognition technology is prospected.
国家科技重大专项(2016ZX05047-002)资助。